LEGO: a novel method for gene set over-representation analysis by incorporating network-based gene weights

Area of Science:

• Bioinformatics
• Computational Biology
• Functional Genomics

Background:

• Pathway or gene set over-representation analysis (ORA) is standard in functional genomics.
• Current ORA tools, like Fisher's exact test, simplify pathways into gene lists, neglecting gene roles and interactions.

Purpose of the Study:

• To develop a novel network-based method, LEGO, for improved ORA.
• To incorporate network-based gene weights, accounting for gene roles and interactions.

Main Methods:

• Developed LEGO (functional Link Enrichment of Gene Ontology or gene sets) incorporating network-based gene weights.
• Evaluated LEGO against Fisher's exact test and other network-based methods in three benchmarks.
• Compared LEGO with gene expression and pathway topology-based methods on 34 disease gene expression datasets.
• Developed a cluster-and-filter approach to reduce redundancy in enriched gene sets.

Main Results:

• LEGO demonstrated superior performance compared to Fisher's exact test and three other network-based methods in benchmarks.
• LEGO ranked among the top methods for sensitivity and prioritization in detecting KEGG pathways across 34 disease datasets.
• LEGO identified relevant autism gene sets missed by Fisher's exact test.

Conclusions:

• LEGO offers a more sensitive and accurate approach to pathway analysis in functional genomics.
• The method enhances the interpretability of results through a novel cluster-and-filter strategy.
• LEGO provides valuable insights into complex diseases like autism by uncovering previously unidentified gene sets.